[BUG] inconsistent behavior of ld -R

[Nick Clifton]

Hi Kristis,

> Thanks for your previous feedback. Unfortunately I've been able to
> reproduce this with ld 2.14.

> H. J. Lu: I'm sorry for not being able to provide a self-contained
> test-case.

Without a testcase it is going to be very hard for us to help you.

> I discovered this while trying to compile an alternate
> implementation of the schedule() function for Linux 2.4.21. You will
> need to have a /usr/src/linux-2.4.21 around with an already compiled
> vmlinux image, since that's the one I tried to use. Other 2.4.x kernels
> may work too.

Why can't you reduce this process to a small, self-contained test case
?  What happens if you try to isolate and reduce the complexity of the
linker's command line ?

Your makefile is passing the contents of a file called "vmlinux" as
the argument to the -R switch.  What is in this file ?  In particular
what are the entries for "runqueue_head" and "aligned_data" ?

Cheers
        Nick